4539

Preliminary Assessment of Intravoxel Incoherent Motion Diffusion-Weighted MRI (IVIM-DWI) Metrics in Alzheimer’s Disease

Maurizio Bergamino¹, Ashley Nespodzany¹, Leslie C Baxter^1,2, Anna Burke³, Richard Caselli², Marwan N Sabbagh⁴, Ryan R Walsh⁵, and Ashley M Stokes¹
¹Division of Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ, United States, ²Mayo Clinic Arizona, Phoenix, AZ, United States, ³Department of Neuropsychiatry, Barrow Neurological Institute, Phoenix, AZ, United States, ⁴Lou Ruvo Center for Brain Health, Cleveland Clinic, Las Vegas, NV, United States, ⁵The Muhammad Ali Parkinson Center, Barrow Neurological Institute, Phoenix, AZ, United States

Synopsis

The objective of this study was to assess complementary metrics from voxel-based morphometry (VBM) and intravoxel incoherent motion diffusion-weighted images (IVIM-DWI) MRI methods in aging populations. Using voxel-based analysis, grey matter (GM) and white matter (WM) differences were analyzed across Alzheimer’s disease (AD), mild cognitive impairment (MCI), and cognitively normal (HC) individuals. IVIM-DWI demonstrated early significant differences between MCI and HC groups, while VBM did not. In addition, voxel-based correlations between neuroimaging metrics and cognitive assessments were assessed in the cognitively impaired groups (AD and MCI).

INTRODUCTION

Alzheimer's disease (AD) is a progressive neurodegenerative disease that affects aging populations. Although structural MRI is widely used to assess AD-induced morphological changes¹, these changes occur late in the AD trajectory and may not be ideal early biomarkers. Functional and microstructural changes precede brain atrophy and may be detectable using advanced MRI in the earlier mild cognitive impairment (MCI) phases. The purpose of this study is to characterize differences in voxel-based morphometry (VBM)², apparent diffusion coefficient (ADC), and IVIM-DWI³ metrics in aging populations. Additionally, the connection between cognitive performance and VBM/ADC/IVIM metrics was investigated.

METHODS

Three aging populations were included in this study: 13 subjects with AD dementia (7 females; mean age±standard deviation (S.D): 77±8 years), 11 subjects with amnestic MCI (8 females; 76±5 years)⁴, and 14 healthy controls (HC, 9 females; 76±7 years). All participants completed the Montreal Cognitive Assessment (MoCA)⁵, Clock-Draw, and the Functional Assessment Staging Tool (FAST)⁶. MRI data were acquired at 3T (Ingenia, Philips). Standard T1-weighted (T1-w) anatomical images were acquired using a MP-RAGE acquisition (TR/TE, 6.7/3.104 ms; acquisition matrix, 256×256; voxel size, 1.06×1.06 mm; slice thickness, 1.2 mm; 170 sagittal slices; flip angle=9°). IVIM-DWI was performed using 7 b-values (25, 50, 75, 100, 200, 500, and 1000 s/mm² with three orthogonal acquisition directions for each b-value; TR/TE, 6000/67.98 ms; acquisition matrix, 96×96; voxel size 2.5×2.5 mm; slice thickness, 2.5 mm). T1-w images were used for GM segmentation using FreeSurfer. All GM-VBM images generated in native space were warped to a T1-w group-wise template space using Advanced Normalization Tools (ANTs, http://stnava.github.io/ANTs/), followed standard procedures. For IVIM, a two-step fitting procedure was performed on the IVIM model³ to obtain D, D*, and f, which represent the pure diffusion coefficient, pseudo-diffusion coefficient, and perfusion fraction, respectively. ADC maps were generated from b=0 and 1000 s/mm² images using a mono-exponential fit. A DWI group-wise template was created from the B0 images, and all ADC and IVIM maps were coregistered to this template through the same procedure described for VBM. Each voxel for VBM, ADC, and IVIM was fit by an ANCOVA model, with gender and age as covariates. All statistical analyses were conducted using the FSL tool Randomise⁷.

RESULTS

Significant differences were detected for the cognitive assessments across groups, as expected. Figure 1 shows the F-values for the significant differences between groups for all analysis methods (VBM, ADC, and IVIM). The subsequent post-hoc analysis, with Bonferroni correction for multiple comparisons, are shown in Figures 2-4 for AD vs. HC, AD vs. MCI, and MCI vs. HC, respectively. Widespread differences were observed between AD and HC, particularly for DWI metrics (Figure 2). More specifically, VBM values were reduced in the AD compared to the HC group in the medial temporal lobe, while DWI identified group differences in insula, parietal, frontal, and temporal lobes, thalamus, caudate, and cerebral WM and cerebral cortex. Comparisons between the AD and MCI groups (Figure 3) revealed differences in the frontal, occipital, parietal, and temporal lobes, right hippocampus, amygdala, and bilateral cerebral cortex using ADC and IVIM metrics. Although VBM identified significant clusters in the temporal lobe and cortex, as well as smaller clusters in the hippocampus and amygdala, these clusters were not observed after the family-wise error (FWE) correction. In addition, no significant clusters were observed between the MCI and HC groups using VBM (Figure 4). On the other hand, significant clusters were detected using ADC and IVIM-D maps in the thalamus, insula, left hippocampus, and amygdala. No significant differences were observed between MCI and HC using IVIM-f. Figure 5 shows the voxel-based correlations between the cognitive assessments and neuroimaging maps.

DISCUSSION

IVIM-DWI metrics may improve sensitivity to sub-voxel microstructural and functional characteristics through the combination of perfusion-insensitive tissue diffusion coefficient and perfusion fraction. We found more significant regional changes using ADC and IVIM-DWI across patients with MCI and AD than standard assessment of GM atrophy with VBM. As expected, the differences were the greatest between AD and HC and occurred mainly in the temporal lobe, hippocampus, and amygdala, which are well-known regions involved in memory-related functions. More specifically, the significant clusters for ADC and IVIM-D covered ~70% of the temporal lobe and > 95% of the hippocampus and amygdala. The differences between MCI and HC were largely located in the thalamus, left amygdala, and left hippocampus using IVIM metrics, which may be consistent with recent evidence of early thalamic involvement in the MCI phase⁸. These results suggest that this method may provide complementary information to standard neuroimaging biomarkers, potentially leading to earlier biomarkers for these changes. Additionally, significant correlations between ACD/IVIM and several cognitive domains were observed, while no significant voxel-based correlations were found with VBM analysis.

CONCLUSIONS

This study demonstrated the potential of IVIM-DWI for assessing AD-related neurodegenerative changes. IVIM-DWI provides complementary information regarding both diffusion and perfusion, which were found to correlate with cognitive decline. These results were compared to standard VBM analysis, which showed good agreement in the later stages of AD. IVIM-DWI may provide early biomarkers of AD, when intervention may prove most beneficial.

Acknowledgements

This work was supported by the Arizona Alzheimer’s Consortium and the Barrow Neurological Foundation.

References

1. Frisoni, G.B., Fox, N.C., Jack Jr. CR, et al. The clinical use of structural MRI in Alzheimer disease. Nat Rev Neurol 2010;6:67–77.

2. Ashburner, J., Friston, K.J. Voxel-based morphometry-the methods. Neuroimage 2000;11:805-821.

3. Le Bihan, D., Breton, E., Lallemand, et al. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology 1988;168:497-505.

4. Petersen, R.C. 2004. Mild cognitive impairment as a diagnostic entity. J Intern Med 2004;256:183-194.

5. Nasreddine, Z.S., Phillips, N.A., Bedirian, V. et al. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J Am Geriatr Soc 2005;53,:695-699.

6. Reisberg, B., Ferris, S.H., de Leon, M.J., et al. The Global Deterioration Scale for assessment of primary degenerative dementia. Am J Psychiatry 1982;139:1136-1139.

7. Winkler, A.M., Ridgway, G.R., Webster, et al. Permutation inference for the general linear model. Neuroimage 2014;92:381-397.

8. Aggleton, J.P., Pralus, A., Nelson, A.J., Hornberger, M., 2016. Thalamic pathology and memory loss in early Alzheimer's disease: moving the focus from the medial temporal lobe to Papez circuit. Brain 139, 1877-1890.

Figures

Voxel-based analysis for the F-statistics across groups for all analysis methods. VBM results are also shown uncorrected for family-wise error (FWE) in panel (a). For all other panels, we report clusters at a significance threshold of p-value < 0.05 corrected for FWE.

Group differences between AD and HC. For VBM analysis we found that the AD group had lower GM volume compare with HCs. For ADC and IVIM maps, the AD group had higher values of ADC, D, and f than HCs. VBM results are also shown uncorrected for FWE in panel (a). For all post-hoc comparisons, the p-values were corrected using the Bonferroni method.

Group differences between AD and MCI. No group differences were observed for VBM analyses with a p-value corrected for FWE. By using the other analysis methods, we found that the AD group had higher values of ADC, D, and f than MCI. Panel (a) shows the VBM results using a p-value < 0.05 uncorrected for FWE. For all post-hoc comparisons, the p-values were corrected using the Bonferroni method.

Group differences between MCI and HCs. Only ADC and IVIM-D detected significant differences for this comparison. For all clusters, we found higher values of ADC and IVIM-D in MCI subjects compared with HCs. For all post-hoc comparisons, the p-values were Bonferroni corrected.

Voxel-based correlations between MoCA/Clock Draw/FAST and VBM/ADC/IVIM metrics. The plots, with the relative mean of R², represent the correlations of the average values inside the significant clusters.

Proc. Intl. Soc. Mag. Reson. Med. 28 (2020)

4539